科研之友
微信
新浪微博
Facebook
分享链接
摘要
An optical fiber pre-warning system (OFPS) can forecast pipeline leakage and prevent serious losses occurring in the field of pipeline transport. Vibration signals of OFPS normally consist of noises, harmful and harmless intrusion signals. With the increase of the surface surrounding complexity along with the buried pipelines, the harmful or harmless intrusion vibration signals will increase seriously. This will be a challenge for the traditional recognition method identifying the harmful intrusion vibration. So it is necessary to add an effective and efficient vibration signal detection link before the traditional recognition method. Only in this way, the noises and harmless intrusion signals can be eliminated effectively, and the data quantity of the late recognition will be decreased greatly. Hence, a new two-dimensional method to detect harmful intrusion vibration for an OFPS is developed in this paper. Two basic features of vibration data, the background homogeneity and the frequent degree of harmless intrusions, are focused to develop the new two-dimensional detection method for OFPS. In the part of space-dimensional detection, two types of Constant False Alarm Rate (CFAR) method, CA-CFAR and GO/SO-CFAR, are used to detect the total vibration data according to the different background homogeneity, respectively. In this way, the huge number of noise signals can be eliminated effectively. In the part of time-dimensional detection, two test methods, Sequential Probability Ratio (SPR) test and Kolmogorov - Smirnov (K - S) test, are adopted to detect the remaining vibration data according to the different frequent degree of harmless intrusions which are obtained by prior information. The harmless intrusion signal data and false alarm data can be eliminated largely. The proposed two-dimensional method is applied to detect the measured field vibration data, which were collected by an OFPS in Rushan of Shandong, China. The analysis results show that the harmful intrusion signals can be detected rapidly and accurately from the complex and huge measured data. Further, the proposed method can ensure a steady false alarm probability in space dimensional detection and improve the detection performance of harmful intrusions effectively.
